Face Detection

This project demonstrates a face detector based on KWIVER using the Haar Cascade classifier. Specifically, this project uses the frontal face classifier. The detector is demonstrated in two languages, C++ and Python. The C++ example implements a detection arrow that offers both image and video options. The Python example is a KWIVER process.

Organization

File Name	Description
face_detection.pipe	Sprokit pipeline file that runs the face detector on a list of input images
face_detection_images.txt	Text file that lists the input images' locations
face_detection.cxx	Face detector written in C++
face_detection.h	Face detector header file written in C++
CMakeLists.txt (C++)	Face detector source and header files added
images	Directory where the sample images are stored
videos	Directory where the sample videos are stored
haarcascade_frontalface_alt.xml	The Haarcascade classifier used
face_detection_video.pipe	Sprokit pipeline file that runs the face detector on a video
faceDetection.py	Face detector written in Python
face_detection_python.pipe	Sprokit pipeline file that runs the python face detector on a list of input images
CMakeLists.txt (Python)	Face detector source file added

Building

In a bash terminal in the project directory::

cmake .


make -j7

Running

C++

To use an image, in <path/to/kwiver/build>/release/bin::

./pipeline_runner -p ../examples/pipelines/face_detection.pipe

After it has finished running, the annotated image will be in ../release/examples/pipelines/output

To use a video, in <path/to/kwiver/build>/release/bin::

./pipeline_runner -p ../examples/pipelines/face_detection_video.pipe

After it has finished running, the annotated frames will be in ../release/examples/pipelines/face_detection_output

Python

In <path/to/kwiver/build>/release/bin::

./pipeline_runner -p ../examples/pipelines/face_detection_python.pipe